Re: [PATCH v2 0/3] nfsd: add a usermodehelper upcall for client id tracking

["J. Bruce Fields" <bfields@fieldses.org>]

On Wed, Oct 03, 2012 at 08:40:28AM -0400, Jeff Layton wrote:
> This is a respin of the set with the same title that I sent on Monday
> (Oct 1st). This one has a few bugfixes, and adds some code to allow the
> upcall program to convert from the legacy database to the new one. I'll
> also be posting a respin of the companion nfs-utils set too.
> 
> The idea here is to allow a one-way conversion from the legacy clientid
> tracking code to the new one, which should allow for seamless kernel and
> nfs-utils upgrades.
> 
> In order to handle that, this set also changes the heuristic that the
> kernel uses to decide what client tracker to use. We now must now try
> the usermodehelper upcall first and only fall back to other schemes if
> its initialization fails.
> 
> Note that I'm not 100% sold on the idea to convert from the legacy db
> format. It's doable but somewhat "fiddly" to handle. There are probably
> corner cases where it will fall down, so it may be best not to even try
> to do that. I won't be offended if you decide not to take the last
> patch.
> 
> Comments? Discuss!

So that's take 3 at reboot recovery.  Take 1 is what people actually use
(despite it being ugly), take 2 nobody's using yet--so it could probably
be ripped out fairly quickly.

I don't see anything objectionable on a quick skim.

I'd rather wait a little longer and consider it for the next merge
window.

--b.